Claude Code
Claude Code is Anthropic’s terminal coding agent. In Frontend it runs on both the browser and server runtimes, on every plan.
Start Claude Code
Section titled “Start Claude Code”Click Claude Code in the sidebar’s Agents group. It opens in the left panel with the project as its working directory.
Sign in
Section titled “Sign in”The first time, run /login and choose how to sign in:
- Claude account — use a Claude Pro or Max subscription (or a Team or Enterprise seat). Claude Code usage comes out of your plan’s limits, shared with Claude on the web and desktop.
- Anthropic Console — pay per token with API credits from your Anthropic account.
Frontend saves your Claude login, encrypted, to your Frontend account, so Claude Code is signed in on your other projects and on either runtime. Run /logout to sign out.
| Link | |
|---|---|
| Compare Claude plans | claude.com/pricing |
| Create an API key | platform.claude.com |
| Sign-in options | Claude Code: authentication |
What’s set up for you
Section titled “What’s set up for you”- MCP servers. The project’s MCP servers are written to Claude Code’s config before it starts, so it can query the project’s database, manage media, generate images and use connected integrations right away. Run
/mcpto see them. - Sandbox instructions. A global
CLAUDE.mdtells Claude Code how the sandbox works, for example that the dev server is already running and how to install packages. - Conversation history. Each project keeps its Claude Code sessions. Run
/resumeto pick one up.
Project instructions
Section titled “Project instructions”Add a CLAUDE.md to the project’s root with conventions Claude Code should follow, such as commands to run, code style or architecture notes. See Claude Code memory.